Riksarkivet, Stockholm · Chifferklaver låda II, no. II:104 · DECODE record R4306
A cipher letter to Amsterdam, read
A one-page letter written entirely in numbers is kept in the Swedish National Archives and is listed in the DECODE database of historical ciphers as non-decrypted. It is a short political letter in Dutch, and it can now be read.
Published 3 October 2026, reviewed and approved by Thomas Leeming. The Dutch text was read by AI models. No specialist in early modern Dutch has checked it yet; corrections are welcome.
The key
Each letter of the alphabet is one number, in alphabetical order, starting at 30.
- There is no j, and u and v share one number. Both are normal for the time.
- Numbers below 30 are blanks. They mean nothing and are there to mislead.
- The numbers 13 and 28 are also used once each as real numbers, for days of the month. The writer marks them with a small raised sign so they are not taken for blanks.
- 57 67 and 66 68 appear once each, as pairs, in the middle of a word. They behave as blanks too.
- 101, 105, 106, 113 and 114 stand for people or powers. The key to these is not known, so they stay as numbers in the text.
The alphabet has nothing that can be adjusted to make a text come out. Either the numbers on the page give Dutch under it or they do not. Three small rules were added from the decoded text: that 13 and 28 with the raised sign are real numbers, that the pairs 57 67 and 66 68 are blanks, and that one group first read as 131 is a stroke followed by 31.
What the letter says
Everything in your letter understood. I shall come (?) and answer by word of mouth, since the matters are of very great weight, and shall bring with me an answer to your request, together with passports. Here mistrust is beginning between [106] and [114]. The envoys of [105] have both left from here. I know the whole intention of [113], what he has in mind. (One word not understood.) To them I have also given advice. He has gone to the [114]. I rest neither night nor day, whether I am sick or not, and shun no danger.
→ Your two letters received, of 28 February and 13 March. Because of a pressing matter I am staying a little longer; it is one that matters.
A note written along the edge of the address side: "(to be addressed?) [113] not otherwise than in the name of [101]."
"Your" is plural throughout: the writer is answering more than one person. The word translated "mistrust" is iaelousye (jalousie), the political term of the time for suspicion or rivalry between powers.
The Dutch
Alles wt den uwen verstaen. Ic sal mondelinge k‹o›men (?) antwoorden, also de saecken ‹v›an seer groten gewichte syn, ende metbrengen antwoort op ulieden ‹v›ersoeck, met paspoorten. Hier begint iaelousye tussen [106] en [114]. [105] gesanten syn beyde van hie(r) vertrocken. Ic weet [113] syn gantsen voornemen, wat hy en den sin heeft. uel (?) Hen heb ic ooc raet gegeven. Hy es na den [114]. Ic rust nacht noch dach, tsy dat ic sieck ben oft niet, en scou geen peryckel.
→ U.L. twee breven ontfangen, van den 28 february en 13 mar‹t›y. Om een noodige saecke / blyf ic noch wat, die importeert. —
Words divided by us. ‹ › marks our correction of a slip by the writer; ( ) a letter the writer left out. Spelling otherwise as written. Note: te r[..]hten (?) [113] niet dan in [101] name.
How it was read
- The obstacle. The one earlier transcription, by D. Bourdeau, read two different round shapes in the units place as zero. One of them is this writer's 8, a figure with a second loop, like an 8 lying on its side. With every 38 and 48 read as 30 and 40, the letters i and t vanished and no language could be recognised.
- A fresh transcription, made blind. The page was cut into enlarged strips. Two AI readers transcribed all of it independently, one from the top down and one from the bottom up. They were told how this writer forms each digit. They were not told the key, the language expected, or the earlier transcription. They agreed on 462 of 471 positions; four of the nine differences were artefacts of joining the strips.
- Eleven places settled on a second look, each with its reason written down. Only for these was the sense of the Dutch used as evidence.
- The key applied by code. Every one of the 35 lines comes out as Dutch.
- Three AI readers of early modern Dutch then worked on the decoded letters: two independently, and one instructed to attack our reading. All three judged the key correct. Two put about nine words in ten as certain; the sceptic put about five-sixths of the text as secure in sense.
A check that came afterwards
The Dutch readers noticed that the blanks never fall between two words. Counted by code: the letter has 22 runs of blanks, and all 22 sit inside a word. If blanks were placed without regard to words, that would happen about once in 360 tries. Ten of the 16 doubled letters in the text are split by a blank, which hides the tell-tale repeated number of ee or oo. The two transcribers knew nothing of this habit, so it supports both the transcription and the key. One limit: where the words divide is our own reading of the Dutch.
The writer also made the same slip three times, writing a number ten too low: kan for van, kersoeck for versoeck, mariy for marty.
The letter, line by line
468 signs in 35 lines (463 numbers and 5 marks): 457 as both blind readers read them (56 of those doubted by at least one reader, shown with ?), and 11 settled on a second look: 7 numbers, shown with *, and 4 of the marks. A dot is a blank. Numbers in square brackets are outside the alphabet.
Show all 35 lines and the note
| Line | Numbers on the page | Letters by the key |
|---|---|---|
| 1 | 30 40 16 40 34 47 50 48 33? 34 42 49 50 | al·leswtdenuw |
| 2 | 34 42 49 34 46 47 48 30? 34 42 38 32 47* 30 40 | enuerstaenicsal |
| 3 | 41 43 42 33 34 40 38 42 16 28? 36 34 39 40 40? | mondelin··gekll |
| 4 | 41 34 42 30 42 48 50 7 11 43 6 43 46 33 34 42 | menantw··o·orden |
| 5 | 30 40 47? 43 33? 6 34 47 30 34 32 39 34 42 | alsod·esaecken |
| 6 | 39 30 42 47 34 57 67 34 46 36 46 5 7 43 | kanse[57][67]ergr··o |
| 7 | 48 34 42 36 16 26 34 50 38 16 32 37 48 | teng··ewi·cht |
| 8 | 34 47 52 42 34 42 33 34 41 34 48 31 | esynendemetb |
| 9 | 46? 34 42 36 34 42 30? 42 48 50 43 43? 46 48? | rengenantwoort |
| 10 | 43 44 49 40 38 34 33 34 42 39 34 46 | opuliedenker |
| 11 | 47 43? 34 32 39 41 34 48 44 30? 47 44 43 | soeckmetpaspo |
| 12 | 43? 46 48 34 42 37? 38? 34 46 31 34 36 | ortenhierbeg |
| 13 | 38 42 48 38 30 26? 34? 40 43 16 19 49 47 | intia·elo··us |
| 14 | 52 34 48 49 47 16 47? 34 42 106 34 42 | yetus·sen[106]en |
| 15 | 114 105 36 34 47? 30? 42 16 12 48? 11 34 42 | [114][105]gesan··t·en |
| 16 | 47 52 42 31 34 52 33 34 49 30 42 37? 38 | synbeydeuanhi |
| 17 | 34 49 34 46 48? 46 43 32 39 34 42 38 32 | euertrockenic |
| 18 | 50 34 11 13* 34 48 113 47? 52 42 36 30? 42 | we··et[113]syngan |
| 19 | 48 47? 34 42 49 66 68? 43? 11 43 46 42 34 | tsenu[66][68]o·orne |
| 20 | 41 34 42 50 30 48 37? 52 34? 42 33 34 42 | menwathyenden |
| 21 | 47? 38 42 37? 34 11 16 34 35 48 49 34 40 | sinhe··eftuel |
| 22 | 37? 34? 42 37? 34 31 38 32 43 43 32 46 30 | henhebicoocra |
| 23 | 34 48 36 34 11 36 34 49 34 42 37 52? 34? | etge·geuenhye |
| 24 | 47 42 30 33 34 42 114? 38 32 46 49 47? | snaden[114]icrus |
| 25 | 48 42 30 32 37 48 42 43 32 37 33? 30 | tnachtnochda |
| 26 | 32 37 48 47* 52 33 30? 48 38 32 47? 38 34 | chtsydaticsie |
| 27 | 32 39 31 34 42 43 35? 48 42 38 34 48? | ckbenoftniet |
| 28 | 34 42 47? 32 43 49 36 34 11 34 42 44 34 | enscouge·enpe |
| 29 | 46 52 32 39 34 40 → 49 40? 48 50 34? | ryckel → ultwe |
| 30 | 34 31? 46 34 49 34 42 43? 42 48 35 30 42 | ebreuenontfan |
| 31 | 36 34 42 49 30 42 33 34 42 28* (raised mark) 35 34 31 | genuanden 28 feb |
| 32 | 46 49* 30 46 52 34 42 13 (raised mark) 41 30 46 38 52 | ruaryen 13 mariy |
| 33 | 43 41 34 16 34 42 42 43 11 43? 33 38 36 34 47 30 34 32 39 34? | ome·enno·odigesaecke |
| 34 | / 31* 40 52 35 38 32 42 43 32 37 50 30 48 33 38? 34 38? 41 44? 43? 46 48? | / blyficnochwatdieimport |
| 35 | 34 34 46 48* — | eert — |
| note 1 | 48 34 46 [hole] 37 48 23? 34 42 | ter [hole] ht·en |
| note 2 | 113 42 38 34 48 33 30 11 42 | [113]nietda·n |
| note 3 | 38 42 101 42 30 27 11 41 34 — | in[101]na··me — |
What is not settled
| Place | On the page | Reading | Standing |
|---|---|---|---|
| Lines 3–4 | k l l(?) men | komen, "come" | Probable |
| Line 6 | kan | van, "of" | Near certain |
| Lines 10–11 | kersoeck | versoeck, "request" | Certain |
| Lines 14–15 | [106] en [114] [105] gesanten | "the envoys of [105]", or [114] [105] as one title and name | Open |
| Line 21 | uel | not understood; wel is the least bad guess | Open |
| Line 22 | hen | "to them" or "to him" | Open |
| Lines 23–24 | hy es na den [114] | "he has gone to the [114]" | Probable |
| Line 32 | mariy | marty, March | Near certain |
| Note, line 1 | ter[hole]ht·en | te richten, "to be addressed"? | Not read |
| Address | A[en] Mr ?i?ick | the name of the man in Amsterdam | Not read |
Who, to whom, and when
Not known. The writer is away from the people addressed and expects to return and report in person, bringing passports. Two envoys have just left the place where the writer is. The writer knows the plans of a man, [113], who has gone to "the [114]".
The language. Dutch, in the clipped style of business and political letters. The writer was very probably not from Holland: some forms point south (Brabant or Flanders), others east (Guelders, Limburg, the Lower Rhine).
The year. Three lines of evidence point three ways and none is established. A later hand pencilled "1632 ?" on the letter. The spelling is old-fashioned and would also suit the later 1500s. One key of exactly this construction survives in the same archive box, issued to a Swedish resident in Moscow, and a second, seen only as a small image, appears to be of the same kind. DECODE dates both from 1650. Neither is this key.
The key to the names. Not found. Of 156 records from this box in the DECODE database, 21 can be viewed, and a new account unlocked no more; none of the keys we read among them has the alphabet in order from 30. By one search of the Swedish National Archives' own catalogue, which we have not re-checked, no images of the collection are online, and the card index of 1934 that should list the key is in the reading room. In the same box, a German cipher text (II:100) was solved by a Swedish colonel in 1935. This one, II:104, was left unsolved, with the addressee's name blank on its wrapper.
Credit, and what this does not claim
- D. Bourdeau published the first transcription of this letter and observed that the numbers behave like one language. This work started from that observation. The transcription used here was made afresh from the scans.
- Riksarkivet holds the letter. DECODE (de-crypt.org), the database of the DECRYPT project, published the scans and the record. Their images are not reproduced here.
- Thomas Leeming set the aim of reading a message no one had read, turned the project to cipher letters when an earlier line of work drifted from that aim, chose this letter from a shortlist that Claude put forward, decided what could be downloaded and kept, created the DECODE account and passed the archive site's human check so that the locked records and the catalogue could be tried (the account unlocked nothing), reviewed the work and approved publication.
- Claude Opus 5.5 (Anthropic) did the searching, the image preparation, the transcription runs, the decoding code, the Dutch readings and the drafting, under that direction; a Claude sub-agent found the key on 2 October 2026. Every step, including the mistakes, is recorded in a lab notebook.
- Not claimed: who wrote the letter, to whom, in what year, or who the five numbered parties are. A finished edition needs a human reader of early modern Dutch to check the text, and the key to the names.